Python Deep Learning Projects

Training loop

Now that the VGG16 model is all set to be used for training, let's load the train_joints.csv file from the train folder containing the IDs of the cropped and resized images with their joint coordinates. 

Then, split the data into an 80:20 train and validation set by using the train_test_split module from sklearn. We imported it with the other imports at the beginning of this chapter. Since the validation data is small, load all of the corresponding images into memory:

Be mindful of how many validation images you load into memory, as this may become an issue with systems that have less RAM.

# load the train datatrain = pd.read_csv('FLIC-full/train/train_joints.csv', header = None)# split train into train and validationtrain_img_ids, ...